EP3403150B1 - Method for monitoring a machine tool, and controller - Google Patents

Description

The invention relates to a method for monitoring a machine tool, in particular a cutting machine tool, according to the preamble of claim 1.

According to a second aspect, the invention relates to a control for a cutting machine tool, with (a) a process variable detection device which is set up to determine process variable measured values of a process variable as a function of a parameter, and (b) a computing unit which has a includes digital storage.

During machining, for example during milling, the process parameters, for example the torque applied to the milling cutter, change constantly. If a machining process is run through several times because several components of the same type are being manufactured, the result is a characteristic course of the machining size over time. If the machining process is disturbed, for example because a milling cutter has broken or because a workpiece has been incorrectly clamped, the time course of the machining size no longer corresponds to the expected pattern.

From the DE 10 2009 025 167 B3 a generic method is known by means of which errors in the processing sequence can be recognized on the basis of deviations from the expected time course of the processing size. In order to be able to compare the resulting process variable curves with one another even with variations in the workpiece geometry during successive machining operations, the time at which the tool intervenes in the workpiece is calculated. For this purpose, a cross-correlation is carried out using the measured values between these points in time. In this way, points in time are determined in relation to a machine time at which the tool enters or exits the workpiece. A disadvantage of such a system is that, for example, unexpected delays in the program sequence can lead to false alarms.

From the article " The power of outliers (and why researchers should ALWAYS check for them) "by _Jason W. Osborne et al, Practical Assessment, Research & Evaluation, Vol. 9, N. 6, March, 2004, ISSN 1531-7714 it is known that there can be many reasons why outliers arise in data sets. It is also discussed what significance these can have and what this means for the corresponding evaluation.

From the article " More intelligence for machine tools "from J. Brinkhaus, Phi Production Technology Hanover, ISSN 1616-2757 it is known to use statistical evaluations of the available sensor data to calculate a probability that a machining process will run error-free at the relevant point in time. If the calculated probability exceeds a predefined limit, the machine tool is stopped. In such a method, too, it is disadvantageous that a change in the processing speed of a machining program leads with a high degree of probability to an error message.

Methods for monitoring machine tools are carried out automatically, either by the machine control itself or by an external processing unit. The program on the basis of which the procedure is carried out must generally be adapted to a new machine tool to be monitored. The reason for this is that machine tools differ in the number of axes, the tools used and their other structure.

The disadvantage of known methods for monitoring machine tools is that they can be prone to false alarms. If the criteria for an alarm are changed in such a way that false alarms occur less frequently, processing processes that are actually faulty can no longer be recognized with the same probability and / or the same short delay time.

From the DE 60 2005 003 194 T2 a controller for controlling a machine tool is known, which is designed to be learning. The controller has an acceleration determination device by means of which the position of the tool, for example, is determined, the position determined in this way being used to control the machine tool. This is advantageous when the machine tool is not viewed as being stiff at will because in this case the position determined by the machine tool using the drives does not have to match the actual position of the tool.

From the EP 1 455 983 B1 a method for acquiring and evaluating process data is known in which measured values are acquired as a function of the running variable and a scatter band of the measured values is determined from several measured value curves. With such a method, the problem described above can arise that mean values are formed via measuring forces that were recorded at different points in time in the processing sequence. This in turn either increases the risk of false positives or reduces the sensitivity of the monitoring process.

The invention is based on the object of improving the monitoring of machine tools.

The invention solves the problem by a method having the features of claim 1. The invention also solves the problem by a generic control which is designed to carry out a corresponding method.

The solution according to the invention has the advantage that the number of false alarms can be reduced without the detection probability and / or speed being negatively influenced. It has been found that false alarms are often caused by the machine control stopping the machining process for a short time, for example because the programmer has to wait longer than expected for a sufficiently high coolant pressure to be available, or because the program sequence is deliberately delayed.

In known monitoring methods, the real time is used as the parameter. If such a delay occurs, the torque applied to a cutter increases later, which can be interpreted as a cutter breakage. Since in the context of the present invention a parameter is used which always characterizes the progress of the machining process, this case cannot occur. If there is a delay, the parameter does not increase any further.

In the context of the present description, the parameter is selected in particular in such a way that it can be described as the argument of the tool trajectory. The tool trajectory is the parameterized curve along which the tool moves. The running variable could also be referred to as the proper time of the machining process. In the theoretical, idealized case, repetitive machining processes can run completely the same, so that the real time, which is measured from a starting point in time of the machining process, was usually used as a parameter. But this has the disadvantages listed above.

A process variable measured value is understood to mean, in particular, a measured value that characterizes a process variable of the machining process of the machine tool. It is possible, but not necessary, for the process variable measured value to be one-dimensional; it is also possible, for example, for the process variable measured value to be a vector, a matrix or a set.

The determination of process variable measured values of a process variable is understood to mean, in particular, the acquisition of data that describe a process variable. For example, process variable measured values are determined by reading out related data from the machine control system. For example, the process variable is a torque that is applied, for example, to a spindle with which the tool is driven. The tool is, for example, a moving tool such as a milling cutter or a drill. For example, the torque of the spindle is determined from its speed and the current engine power.

The feature that it is determined whether the process variable measured values are within the specified tolerance interval is understood in particular to mean that it is checked whether the course of the process variable measured values lies in a tolerance band. The tolerance band is the result of all tolerance intervals. In other words, the tolerance band is a flat object, whereas the tolerance interval is a linear object.

The feature that a warning signal is output in the event of a negative is also understood, in particular, that no warning signal is output in the affirmative.

In other words, in the normal case that the process variable measured values are within the specified tolerance interval, no warning signal is output.

The feature that the parameter always characterizes the progress of the machining process is understood in particular to mean that the parameter changes when and only when the machining process progresses.

The method comprises the step of calculating the running variables from real time and at least one process parameter that characterizes the processing speed of the machining process. The process parameter is an input variable. In other words, the process parameter is not calculated as part of the method. Rather, the process parameter is recorded externally. For example, the process parameters are read out by the machine control, which can slow down, accelerate or stop the machining process on the basis of the algorithm on which the control is based.

It is particularly favorable if the at least one process parameter is an instantaneous total speed value. The overall speed value can also be referred to as the override value, as the overall speed controller is often referred to as the override controller. With an overall speed controller, the processing speed of the machining program and thus the speed of the machining process can be influenced directly. A total speed value of 1 or 100% corresponds to the speed specified in the machining program. The machining program is the sequence of commands that codes the machining of the workpiece. For example, it is an NC program.

The total speed value is the value that describes the resulting processing speed in relation to the speed specified in the machining program. It is possible that several partial speed controllers exist. In these, only their overall effect is relevant.

If the overall speed controller is set to 110%, for example, the tool, for example the milling cutter, moves 10% faster than if it were set of 100%. It is possible, but usually not provided, for the overall speed controller to influence the speed of the spindle for driving a tool. For example, the real time value that characterizes the position of the tool when the overall speed controller is at 100% and there are no downtimes can therefore be used as the running variable.

If the total speed value is used to calculate the running variable, it is preferably done by numerically calculating the integral over the current total speed value. This integral is represented numerically in that the sum is formed over products, according to which one factor is the time interval and the second factor is the current total speed value in the time interval. When the limit is crossed for any small time interval, the integral over the total speed value then results. It should be noted that the run variable defined in this way also has the dimension second.

In its preferred embodiment, the at least one process parameter comprises a standstill time, which characterizes a standstill of the machining process. Many machine tool controls are designed in such a way that they stop the machining process if predetermined threshold values are not reached, for example cooling lubricant pressure or spindle speed, and / or an axis is not released. This downtime is managed in the program independently of the override value. During the downtime, the machining process does not progress and the running variable does not change accordingly.

The step of determining whether the process variable measured values are within the specified tolerance interval comprises the following steps: (b1) for a run variable for which a process variable measured value was determined, determining a time environment around this run variable, ( b2) determining at least one reference running variable from the time environment for which there is at least one reference process variable measured value that was recorded in a previous, same machining process, and (b3) calculating the tolerance interval from the at least one reference Process variable measured value. This procedure is based on the knowledge that when executing a machining process, for example using a CNC program, the process variable measured values are only recorded in the theoretical ideal case with the same values for the run variables.

Since there are delays in every real sequence of the machining process and these can only be characterized by the run variable within the scope of the numerical accuracy, it can happen that there is no reference process variable measured value for a certain value for the run variable, but there is for a run variable that is only a small distance from the relevant value for the run variable. Therefore, for a given value of the run variable in the time environment around this value of the run variable, a search is made for reference run variables for which a reference process variable measured value exists.

Of course, the time environment must not be too large, since otherwise the tolerance interval will be calculated too large. It is beneficial if the time interval is less than 0.5 sec.

The tolerance interval is preferably calculated using a maximum and a minimum using the reference process variable measured values (B _ref (i _ref ). This is understood in particular to mean that the interval limits are calculated using a formula that contains the maximum and the minimum. It is possible, but not necessary, for the formula to contain other variables, for example a measure of dispersion.

Alternatively, the tolerance interval is calculated from a mean value and a degree of dispersion of at least two reference process variable measured values. The mean value can be the arithmetic mean value, for example. Alternatively, the mean can also be a truncated mean, a winsorized mean, a quartile mean, a Gastwirth-Cohen mean, a range mean or a similar mean. The measure of dispersion can be the variance or the standard deviation. However, it is also possible that, for example, a capped variance or a capped standard deviation is used.

According to a preferred embodiment, the method comprises the steps of detecting an end of a positioning movement and / or a start of a feed movement and setting the running variable to a predetermined value, when the end of the positioning movement and / or the start of the feed movement have been detected. In a program, in particular a CNC program that codes a machining process, in most cases positioning movements and feed movements can be distinguished from one another. The aim of a positioning movement is to bring the tool to a specified position without it being in the workpiece. Positioning movements are usually carried out with the highest possible axis speed in order to keep the processing time as short as possible.

A feed movement, on the other hand, is only carried out at such a high speed that the tool and / or the workpiece are not overloaded. During the feed movement, the tool is engaged or moves before or after the engagement with the workpiece at the same speed as during engagement. Since numerical errors can occur when calculating the run variable, it is advantageous to set the run variable to a previously defined value when an easily identifiable point in time has been reached in the machining process. The end of a positioning movement or the start of a feed movement are well suited for this purpose.

A cascade controller is preferably implemented in a controller according to the invention. A cascade controller is understood to mean a controller which has a plurality of control loops, the higher-level controller in each case providing the setpoint value to the downstream controller. For example, the controller of the highest hierarchy level can be a position controller that controls a target position of the tool. Deviations between target and actual positions and the time available for adjustment lead to a target speed that is regulated by a hierarchically subordinate speed controller.

This can be followed by a torque regulator which also regulates the target torque, which results from the difference between the target speed and the actual speed. A current regulator can in turn be connected downstream of this, which in turn drives a voltage regulator. The deeper the hierarchy level, the higher the frequency with which the controller works. For example, the position controller has a frequency between 50 and 500 Hz, whereas the current controller can already have a frequency between 5 and 15 kHz. The cascade controller is preferably controlled by an NC program controlled, which is stored in the digital memory and encodes the machining process.

Figur 1

eine schematische Ansicht einer erfindungsgemäßen Werkzeugmaschine zum Durchführen eines erfindungsgemäßen Verfahrens,

Figur 2

einen Prozessgrößenverlauf,

Figur 3

eine schematische Ansicht von drei unterschiedlichen Prozessgrößenverläufen, die zu unterschiedlichen Wiederholungsindizes gehört,

Figur 4

eine Veranschaulichung der Messwert-Menge und

Figur 5

den Erwartungswertverlauf des Bearbeitungsprozesses.

The invention is explained in more detail below with reference to the accompanying drawings. Show:

Figure 1

a schematic view of a machine tool according to the invention for performing a method according to the invention,

Figure 2

a process variable curve,

Figure 3

a schematic view of three different process variable courses belonging to different repetition indices,

Figure 4

an illustration of the amount of measured values and

Figure 5

the expected value curve of the machining process.

Figure 1 shows schematically a machine tool 10 with a tool 12 in the form of a drill. The tool 12 is driven by a schematically drawn spindle 14. A workpiece 16 is clamped on the machine tool 10 and is machined by the tool 12 as part of a machining process.

The spindle 14 and thus the tool 12 can be positioned in three spatial coordinates, namely in the x-direction, y-direction and z-direction. The corresponding drives are controlled by an electrical controller 18 which includes a digital memory 20. A CNC program is stored in the digital memory 20. A program for carrying out a method according to the invention is also stored in the digital memory 20 or a digital memory that is spatially separated therefrom.

The machine tool can also have a schematically drawn sensor 22, for example a force sensor or an acceleration sensor, which measures the acceleration of the tool 12 or the spindle 18 or another component or a force acting on such a component.

To carry out a machining process, the controller 18 processes the CNC program stored in the digital memory 20. In this program, positions to be approached with tool 12 and speeds for its movement are stored. The controller 18 calculates a trajectory from this r ( i ) = ( x, y, z ) ( n ) as a function of a program step counter n from a specified starting time. At the end of the program, the controller 18 moves the tool 12 back to the starting point. At the start of such a machining process, the program step counter is reset, for example to the value n = 0.

At the end of the machining process, the workpiece 16 is removed and replaced by a new, identical workpiece, so that the same machining process is carried out again. The process in which two holes are made in the workpiece 16 is considered below. The position in which the second hole is made is shown by the tool and spindle drawn in dashed lines.

In this case, a machining process comprises the positioning of the tool 12 at the first position r ₁ = (x ₁ , y ₁ , z ₁ ), drilling the hole, moving the tool 12 out of the workpiece 16, positioning it in the second position r 2 = (x ₂ , y ₂ , z ₂ ), drilling the second hole, moving the tool 12 out of the workpiece 16 and moving back to the starting position.

During this machining process, a drive torque M _A , which the spindle 14 applies to the tool 12, is repeatedly detected by the controller 18. Alternatively, there is a computing unit which is independent of the controller 18 and which reads _{the drive torque M A from the controller 18.}

From any position r ₁ , r 2 from the tool 12 is moved into the workpiece 16. The position at which the tool 12 first touches the workpiece 16 has the z coordinate z _beginning and the position at which the tool 12 engages maximally deep into the workpiece 16 has the z coordinate z _end . The positions differ for each hole because the x coordinates are different, apart from any differences in thickness of the workpiece 15, however, the respective z coordinates are the same.

Figure 2a shows schematically the process variable curve B ₁ (n) = M _A (n) for an ideal machining process. This process variable curve plots the determined drive torque M _A against the program step counter n. In the ideal case, the progress by one program step counter always corresponds to the same time interval Δt. It can be seen that the process variable M _A begins to increase at n = 3 sec. This is the point in time at which the drill 12 comes into engagement with the workpiece 16. So we have z = z _beginning . In Figure 2a a program step counter corresponds to a time interval of real time of 0.1 sec (sec = seconds).

At the end of the drilling process, the drill 12 is pulled out of the drilled hole, and the drive torque M _A falls when z = z _end . Then the drill 12 is repositioned and a further hole is drilled, the drive torque M _A increasing again from t = 30 sec when z = z _beginning applies.

Figure 2b shows the case in which the machining process is ideal for the first hole. After machining the first hole, however, there is a standstill time Δt _still , which lasts two program step counters. In addition, after the first hole, an overall speed controller or override controller 23 (see Figure 1 ), which reduces the overall speed, which could also be referred to as processing speed or running speed, to 70% of the original speed. This can be done, for example, to reduce tool wear.

Both lead to the fact that, for example, at time t = 4 sec in the second run, a process variable B ₁ = M _A = is significantly smaller than at time t = 4 sec in the first run.

Figure 2b shows, in addition to the time scale of real time t, a scale with a running variable i, which of course does not correspond to the imaginary unit. The running variable i is a real number in the idealized case. The running variable i in the present embodiment is calculated as $$i\left(t\right)={\displaystyle \underset{t\prime =t_0}{\overset{t\prime =t}{\int }}O\left(t\prime \right)\mathit{German}\prime -{\displaystyle \sum \mathrm{\Delta }{t}_{\mathit{quiet}}}}$$

In other words, downtimes in the machining process also lead to a standstill of the running variable i. If the total speed value O is less than 1, the real time t is weighted and integrated.

It should be added that the running variable i can of course also be calculated, for example, by setting the total speed value O (only) to zero when calculating the integral during idle times. Other types of calculation are also possible, but for each of them it applies that idle times do not lead to the running variable i becoming larger.

It can be seen that the run variable i has the dimension time. In the idealized but not real case, i.e. without downtimes and always unchanged processing speed, i (t) = t + to applies, where t _{0 is} the respective starting time of the machining process.

If two machining processes run smoothly, the tool is in the same place relative to the workpiece for every value of the running variable i, apart from numerical errors, even if there are downtimes or a change in processing speed. Numerically, Formula 1 is represented by a sum.

The Figure 3 shows three courses of process variable measured values, namely B ₁ (i), B ₂ (i) and B ₃ (i), the subscript being the repetition index k. Every time the tool 12 machines a new workpiece 16, the repetition index k is increased by one. A numbered set of process variable curves B _k (i) = M _k (i) is thus obtained. The current machining process is the one with the repetition index k = 3.

Figure 4 shows two courses of process variable measured values for k = 1 and k = 2. The processing process with the repetition index k = 3 is almost complete, the last recorded process variable measured value is (B (45)) for the run variable i = 45.

To determine whether the process variable measured values (B (i = 45)) lie within a predetermined tolerance interval T (i = 45), a time environment U _e (45) is first determined, with the size e of the environment is selected, for example, such that the tool has covered a predetermined distance during the period described by the environment, this distance preferably being at least 500 μm and at most 5000 μm. In the example, e = i, so that all process variable measured values B _k (44), B _k (45) and B _k (46) for k = 1 and k = 2 are in U ₁ (45). The fact that all i in the present example are whole numbers serves to simplify matters, in the real case the i do not have to be whole numbers.

The expected value E (45) is the mean value from the reference process variable measured values B ₁ (44), B ₁ (45), B ₁ (46), B ₂ (44), B ₂ (45) and B _{2 (46)} and the variance σ ² (45) is calculated as the degree of dispersion and from this the tolerance interval T (45) = {E (45) - σ ² (45); E (45) + σ ² (45)}. This calculation is carried out for all i of the current machining process. In Figure 4 the calculation is also shown for i = 18.

It is possible, but not necessary, that not all B _k (i) that lie in the time environment U _e (i) are used to calculate the tolerance interval. With a large number of repetitions, it can make sense that the repetition set contains the last twenty repetition indices, for example, in order to keep the calculation small.

Figure 5 shows the expected value curve E (i) after a large number of good machining processes, i.e. machining processes that were processed without errors. In Figure 5 the tolerance interval T (45) is also shown purely schematically. The area between the dashed curves is the tolerance band.

List of reference symbols

10

Machine tool

12th

Tool

14th

spindle

16

workpiece

18th

control

20th

digital storage

22nd

sensor

24

Allowance fluctuation

r (i)

Trajectory

i

Run variable

MA

Drive torque

k

Repetition index

n

Program step counter (natural number)

T

Tolerance interval

t

real time

U

Surroundings

O

Total speed value

Claims (10)

1. A method for monitoring a machine tool (10), in particular a material-removing machine tool (10), having the steps of:

   (a) determining process variable measured values (B(i)) of a process variable (B) on the basis of a parameter,

   (b) determining whether the process variable measured values (B(i)) lie within a predefined tolerance range (T(i)) which depends on the parameter,

   (c) if not, outputting a warning signal and

   (d) constantly repeating steps (a) to (c),

   characterised by the step:
   (e) calculating the parameter as the argument of the tool trajectory from the real time (t) and at least one process parameter (O, Δt_still) which characterizes the processing speed of the machining process, so that the parameter is a control variable (i) which always characterizes a progress of the machining process.
2. The method according to claim 1, characterised by the fact that the at least one process parameter (O, Δt_still) is a momentary velocity value (O(t)) of an overall velocity regulator.
3. The method according to claim 1 or 2, characterised by the fact that the calculation of the control variable (i) from the real time (t) comprises the calculation of the integral over the momentary overall velocity value (O(t)).
4. The method according to one of the above claims, characterised by the fact that the at least one process parameter (O, Δt_still) comprises a downtime (Δt_still) that characterizes a stationary point in the machining process.
5. The method according to one of the above claims, characterised by the fact that the step (b) of determining whether the process variable measured values (B(i)) lie within the predefined tolerance range (T(i)) includes the following steps:

   (b1) for a control variable (io) at which a process variable measured value (B(io)) has been determined, determining a time neighbourhood U_e(i₀) around this control variable (io),

   (b2) determining at least one reference control variable (i_ref) from the time neighbourhood U_e (io), for which at least one reference process variable measured value (Bref(iref)) exists which has been recorded in a previous, identical machining process, and

   (b3) calculating the tolerance range (T(io)) using the at least one reference process variable measured value (B_ref(i_ref)).
6. The method according to one of the above claims, characterised by the fact that the tolerance range (T(i)) is calculated using a maximum and a minimum above the reference process variable measured values (B_ref(i_ref).
7. The method according to one of the above claims, characterised by the step:

   (a) recording an end of a positioning movement and/or a start of a feed movement, and

   (b) setting the control variable (i) to a predefined value.
8. Controller (18) for a material-removing machine tool (10) with

   (a) a process variable recording device that is configured to determine process variable measured values (B(i)) of a process variable (B) based on a parameter, and

   (b) a processing unit which comprises a digital memory (20),

   characterised by the fact that
   a programme code is saved in the digital memory (20) that codes a method according to one of the above claims.
9. Controller according to claim 8, characterised by the fact that a cascade regulator is implemented within it.
10. Machine tool (10) with a controller (20) according to claim 8 or 9.

Images